A two-level neural network approach for flicker source location

Haidar Samet (Corresponding author), Mahdi Khosravi, Teymoor Ghanbari, Mohsen Tajdinian

Research output: Contribution to journalArticleAcademicpeer-review

17 Downloads (Pure)


Identification of flicker sources is necessary to find who is responsible for the measured flicker and improve power quality. This paper puts forward a new method for identifying flicker sources with minimum measurement units. Contrary to the previous works where flicker sources are considered a single-frequency signal, the autoregressive moving average (ARMA) is used to model active and reactive power variations. First, the envelope of the network voltage at the considered busbars is derived by the Hilbert transform. Then, appropriate flicker indices are extracted from the power spectral density (PSD) of the voltage envelope. A novel two-level structure of a set of ANNs is proposed, which needs a low number of voltage measurement units to locate the flicker sources. Using the captured data from different simulations of various scenarios, the Artificial Neural Networks (ANNs) are trained to categorize flicker sources.

Original languageEnglish
Article number107157
Number of pages17
JournalComputers and Electrical Engineering
Publication statusPublished - Jun 2021


  • ANN
  • Autoregressive Moving Average (ARMA)
  • Flicker source identify
  • Hilbert transform
  • Power quality


Dive into the research topics of 'A two-level neural network approach for flicker source location'. Together they form a unique fingerprint.

Cite this